KM ^3 SVM: A Efficient Min-Max Modular Support Vector Machine Based on Clustering

In recent years, more and more scholars have begun to study the problem of multi-label learning. In this paper, we propose a multi-label learning method called KM\(^{3}\)SVM based on Clustering idea. First, for each label, KM\(^{3}\)SVM method selects positive samples and negative samples, then uses clustering method to gain a number of training subsets. By using these relatively smaller and more balanced training subsets, we can get a group of classifiers. Given an unseen sample, we can obtain a series of outputs and combine them by two simple principles. Experimental results on two datasets demonstrate the superiority of the proposed method KM\(^{3}\)SVM over several state-of-the-art multi-label learning algorithms.